AIX regression due to DFA scheduler merge

[David Edelsohn]

>>>>> law  writes:

Jeff> This makes me wonder if fixing the patch may be as simple as moving up
Jeff> the call to create the sched groups.  That would presumably make 
Jeff> add_dependence work as expected, which in turn would cause the code which
Jeff> turns backwards dependencies into forwards dependencies to work as expected.

	We can try, but Michael Matz and I were concerned about
interfering with the other information collected during sched_analyze if
we set SCHED_GROUP_P earlier.

Thanks, David